<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<p>Hi Havard,</p>
<p>Accepting sparsely TSIG signed transfers is a requirement of RFC
8945 (see
<a class="moz-txt-link-freetext" href="https://www.rfc-editor.org/rfc/rfc8945.html#name-tsig-on-tcp-connections">https://www.rfc-editor.org/rfc/rfc8945.html#name-tsig-on-tcp-connections</a>
):</p>
<p> "For backward compatibility, a client that receives DNS
messages and verifies TSIG <span class="bcp14">MUST</span> accept
up to 99 intermediary messages without a TSIG and <span
class="bcp14">MUST</span> verify that both the first and last
message contain a TSIG."</p>
<p>Also, CVE-2026-19033 is about the issue that named didn't roll
back to the state before a transfer was started when an invalid
TSIG is occurred (see: <a class="moz-txt-link-freetext" href="https://app.opencve.io/cve/CVE-2026-19033">https://app.opencve.io/cve/CVE-2026-19033</a> ).
Requiring a TSIG on every message is a radical remedy (breaking
the standard) and could have been addressed more on-target
instead.</p>
<p>That said, RFC 8945 does require servers to TSIG sign all the
outgoing messages (Postel's law), so I'll have a look if I can
provide a patch for OpenDNSSEC to do just that.</p>
<p>I'll keep you posted.</p>
<p>Best regards,</p>
<p>-- Willem</p>
<div class="moz-cite-prefix">On 9/18/26 14:55, Havard Eidnes via
Opendnssec-user wrote:<br>
</div>
<blockquote type="cite"
cite="mid:20260918.145528.1314790663464938093.he@uninett.no">
<pre wrap="" class="moz-quote-pre">Hi,
it appears that one of the security fixes in the BIND 9.20.29
release is hitting a possible bug of omission in OpenDNSSEC, at
least in version 2.1.13 (no, I'm not on the latest...), ref.:
Require a TSIG on every message of incoming zone
transfers. (CVE-2026-19033)
Previously, named accepted TSIG-signed zone transfers in
which some messages were unsigned, and processed those
messages before the next signature could vouch for them. It
now requires a TSIG on every message of an incoming AXFR or
IXFR; all modern nameservers already sign every message, so
no change is expected in practice. [GL #6062]
After upgrading the "downstream" name server of OpenDNSSEC to
BIND 9.20.29, I'm seeing
Sep 18 14:26:33 xxxxx ods-signerd: [socket] unable to handle outgoing tcp response: write() failed (Broken pipe)
messages logged by OpenDNSSEC, and on the BIND side:
Sep 18 12:05:21 yyyyy named[22162]: 0x6fe6d25cec10: transfer of '114.39.158.in-addr.arpa/IN' from a.b.c.d#53: failed while receiving responses: expected a TSIG or SIG(0)
(No, those are not the same event, but the pattern is clear.)
For now I'm downgrading the downstream BIND name server to the
previous BIND release, 9.20.27 (28 was skipped).
It would however be nice if this bug of omission in OpenDNSSEC
could be fixed. Any chance of that happening?
Best regards,
- Håvard
_______________________________________________
Opendnssec-user mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Opendnssec-user@lists.opendnssec.org">Opendnssec-user@lists.opendnssec.org</a>
<a class="moz-txt-link-freetext" href="https://lists.opendnssec.org/mailman/listinfo/opendnssec-user">https://lists.opendnssec.org/mailman/listinfo/opendnssec-user</a>
</pre>
</blockquote>
</body>
</html>